Research Of The Segmentation Technology For VCH-F2Abdominal Slice Image

Virtual Human Project, a very significant research topic widely used in so many fields such as medicine, aerospace, bionics and further self recognition of human beings, aims at constructing "a human being" on computer by way of information processing technology.It includes four stages in development:Virtual Visual Human; Virtual Physical Human; Virtual Physiology Human and Virtual Intelligence Human.The study of this Project now in our country runs at the first stage, namely the process of3D reconstruction and understanding of body and tissue organ.The medical image segmentation method plays one of the most important roles at the Virtual Visual Human stage, while the key step for this stage relies on the successful segmentation from tissue organ of color human body slices image.This thesis, on the basis of detailed analysis and study for the basic theory of image segmentation, put its emphasis on the region-based, boundary-based segmentation method and image segmentation method of a combination of specific theoretical tools.Secondly,with references of the abdominal slice data image characteristics from Virtual Chinese Human-Female NO.2(VCH-F2) offered by the Third Military Medical University, image pretreatment was carried out. This process mainly includes three steps:gray color slice images, image format conversion, cropping and filtering. The purpose of these pre-processing is to remove the image noise, to improve image signal-to-noise ratio, and make necessary preparations for image segmentation work afterward.Thirdly, This thesis, given detailed analysis and discussion of the traditional segmentation algorithm for live-wire, and of the deficiencies as well as its causes of such segmentation algorithm, made improvement in two aspects of cost function and optimal path search.Improvement of cost function mainly focuses on the replacement of Canny operator edge detection characteristic function for Laplace edge detection characteristic function of the cost function, and applies non-linear mapping (f (x)=exp(-4x2)) to reconstruct cost function. Optimal path search joint stop searching conditions, namely it indicated that the optimal path from the starting point to the target point had been found when the target node was added to the set of nodes of the shortest path, then you can stop the trail search, and followed by reverse lookup you can get the optimal path direction, As a result you can greatly save the path search time and improve the computational efficiency of the algorithm.Experimental results show that the segmentation method given in this research can accurately cut the human kidneys from abdominal slice image only with less human involvement. And compared with the original segmentation algorithm, it gets improvement both in the accuracy of segmentation algorithm and operating efficiency.